Neuromorphic computing has attracted great attention to mimic the brain functions of perception, learning and memory, which are considered to overcome the von Neumann bottleneck. Here, we developed a novel neuromorphic device based on the ZnO/PbS hybrid heterostructure in order to emulate the bio-synaptic activities in the fully photon modulated mode. Owing to regulation of the conduction state, the excitatory and inhibitory activities have been emulated with the excitation of long-wavelength and short-wavelength photons in the ZnO/PbS neuromorphic device. Excitatory plasticity can be mimicked with the UV light, and IR light induces the inhibitory effect. Furthermore, other synapse functions have also been emulated in these modes, including long-term plasticity, short-term plasticity, paired-pulse facilitation/depression, spike-rate-dependent plasticity, etc. Meanwhile, an artificial neural network has been simulated based on the synaptic plasticity of the excitatory and inhibitory effect in the fully photon modulation, and recognition rates up to 67 ± 6% can be achieved to distinguish the letter images. Our multifunctional artificial synapse based on the fully photon stimulation can open up a new and efficient way for constructing neuromorphic systems.
中文翻译:
神经形态计算已引起人们极大的关注,以模仿感知,学习和记忆的大脑功能,这些功能被认为可以克服冯·诺依曼瓶颈。在这里,我们开发了一种基于ZnO / PbS杂化异质结构的新型神经形态设备,以便在完全光子调制模式下模拟生物突触活动。由于调节了传导状态,在ZnO / PbS神经形态装置中通过长波长和短波长光子的激发来模拟兴奋和抑制活性。紫外线可模拟兴奋性可塑性,而红外线则具有抑制作用。此外,在这些模式下还模拟了其他突触功能,包括长期可塑性,短期可塑性,成对脉冲促进/抑制,同时,基于神经元在全光子调制中的兴奋性和抑制性作用的突触可塑性模拟了一个人工神经网络,可以识别高达67±6%的识别率。字母图像。我们基于全光子刺激的多功能人工突触可以为构建神经形态系统开辟新的有效途径。